Sviatlana Karpava
1,4, Maria Kambanaros
2,4, Kleanthes K. Grohmann
3,4[email protected], [email protected], [email protected]
1University of Central Lancashire, Cyprus, 2Cyprus University of Technology,
3University of Cyprus, 4Cyprus Acquisition Team
Abstract. Narratives can help identify linguistic, cognitive, semantic, and social abilities as well as the communicative competence and cultural awareness of a child; cultural communities, language environment, home language use, parental attitudes towards bilingual and bicultural learning, and level of language proficiency are some of the factors that influence the development of narrative abilities. The pilot study reported here investigates narrative performance by Russian- (Cypriot) Greek bilingual children in both of their languages, Russian and Cypriot Greek. A total of 23 simultaneous bilingual children across different age groups ranging from 3 to 11 years of age were tested with the MAIN, the Multilingual Assessment Instrument for Narratives, a narrative tool developed within COST Action IS0804. All participants were also tested on a battery of other tools, including proficiency tests for Russian and Cypriot Greek as well as several tasks assessing executive control. With regard to narrative abilities, the bilingual children performed similarly across both languages. Not surprisingly, their performance was higher on the retelling condition than on the telling condition. Also as expected, the bilingual children’s narrative abilities improve with age, although the number of participants in each age group is too low to determine a concrete trajectory. A comparison of the bilingual participants’ telling and retelling narrative productions with that of monolingual (standard) Greek- and monolingual Russian-speaking peers shows that these outperform the bilinguals mainly in story structure and internal state terms.
Keywords: bilingualism, narrative, communicative competence, macro-structure, telling, retelling
Introduction
The present study investigates the narrative performance of bilingual children with typical language development in both their languages, Russian and Greek. Concretely, as the research takes place in Cyprus, which is characterised by diglossia between the local variety and the standard language (see Rowe & Grohmann, 2013 for a recent overview), Cypriot Greek (henceforth, CG) was assessed where relevant. The relevance is three-fold. First, the local variety (CG) spoken in a linguistic environment where the official language is Standard Modern Greek (SMG) leads to children growing up to become
‘(discrete) bilectal’ speakers (Rowe & Grohmann, 2013); for narrative abilities, it would be interesting to be able to distinguish between monolingual-mono(dia)lectal and monolingual-bilectal children.
Second, it will be instructive to compare bilingual-mono(dia)lectal children with those participating in our research, bilingual-bilectal ones; they are arguably simultaneously bilingual, yet sequentially bilectal (for discussion, see Grohmann & Kambanaros, to appear). Third, the data of our particular group of Russian–Greek children, namely bilingual-bilectals, can be compared to that of their peers acquiring either language monolingually (regardless of dialectal issues); these would be children from Russia and Greece, respectively. (A fourth possible relevance is briefly presented right below.) For the purposes of this research, narrative performance is measured by macro-structure in telling and retelling conditions, along the dimensions of story structure, structure complexity, and internal states terms. Other factors that have already been partially considered (and will be expanded in the future), include children’s language competence, language of narration, executive control, chronological age, and schooling level, which have all been identified as relevant in the bilectal context (Grohmann &
Kambanaros, to appear).
We choose narratives as a window into the bilingual children’s communicative development because it has been argued that narrative performance can help identify linguistic, cognitive, semantic, and
126
social abilities as well as a child’s communicative competence and cultural awareness (e.g., Olley, 1989; Schneider, Hayward, & Dubé, 2006; Paradis, Genesee, & Crago, 2010). In fact, narrations are often employed in contexts of language assessment, for example, to probe for possible language delay or impairment (e.g., Leonard, 1998); as a fourth possible relevance of our research, then, data from language-impaired children could be used for additional comparison. In the long run, this becomes particularly relevant in the context of early diagnosis of speech/language and communication difficulties in bilingual children who often show language behaviour that is reminiscent of language- impaired children, even and especially in the absence of any developmental language problems (for a very recent up-to-date overview of these issues, see Armon-Lotem & de Jong, 2015).
It has also been suggested that narrative skills are important for children’s scholastic achievements, since there is a close relationship between oral language skills and literacy (Snow, 2002). In turn, cultural communities, language environment, home language use, parental attitudes towards bilingual and bicultural learning, and the level of language proficiency are some of the additional factors that may affect children’s development of narrative abilities (e.g., Jia, Yiu, Duncan, & Paradis, 2011).
Narratives are thus one measure to assess children’s speech and language abilities (e.g., Hadley, 1998;
Boudreau, 2008) and their communicative competence (e.g., Norbury & Bishop, 2003). Analysis of narrative productions can also be applied to different cultural and social populations such as bi- and multilingual children. These differ from their monolingual peers regarding narrative abilities, with some relevant dimensions of evaluation being dissimilar languages, possible variance in proficiency levels for the two or more languages, a different language environment, and more diverse cultural communities (Gutiérrez-Clellen, 2002; Fiestas & Peña, 2004). Narrative micro- and macro-structures also depend on non-linguistic factors such as executive control, which comes out as the effect of working memory, sequencing, and planning on the stories produced (Coelho, 2002). Note that there is a long-standing claim that bilingualism enhances children’s development of executive functions, that is, the set of cognitive processes that underlie flexible and goal-directed behaviour, often referred to as the ‘cognitive advantage of bilingualism’ (for recent overview, see Costa & Sebastián-Gallés, 2014).
Turning to analyses of narrative productions, the data reported here have not yet been subjected to a detailed micro-structural analysis, which is sensitive to language-specific aspects; macro-structure is less language-specific and even language-dependent (e.g., Uccelli & Paez, 2007), which is why we concentrate on this level of analysis for the present study. Macro-structural story structure has been analysed to include story grammar, narrative quality, plotline, temporal-causal connection, episode structure completeness, and so-called Goal-Attempt-Outcome structure. Story grammar itself requires knowledge of both semantic-pragmatic information and a super-structural level of discourse organisation (see Gagarina, Klop, Kunnari, Tantele, Välimaa, Balčiūnienė, Bohnacker, & Walters, 2012; 2015 for details and references).
Method
Research questions
The main research questions of this study are the following:
1. With respect to narratives, do bilingual-bilectal children perform differently in each of their languages, Russian and (Cypriot) Greek?
2. Does mode of narration (telling/retelling) influence story structure, structural complexity, and the production of internal state terms by bilingual children, in either language or even both?
3. Are the bilingual children’s narrative productions similar to or different from monolingual children, language-impaired children, and other bilinguals with different language pairs?
4. What role do variables such as age, schooling, proficiency level, cognitive abilities, and executive functions play in bilingual children’s narrative performance?
What we report next is the result of a pilot study. It is meant to pave the way for a larger-scale cohort research project, which is why we chose (few) participants for all age groups ranging from as low as 3 to as old as 11 years of age. One rationale is, of course, to test the validity of the tool used, the MAIN
Proceedings ISMBS 2015
127
(Gagarina et al., 2015): Does it work for this particular bilingual population of children acquiring Russian and (Cypriot) Greek? Can it differentiate bilingualism (Russian, Greek) from bilectalism (Cypriot Greek, Standard Modern Greek)? Is it age-appropriate across a wide range? Does it produce sound data that can be analysed quantitatively and qualitatively? From this perspective, a different way to present the data would be as many different single-case studies, namely one for each of the 23 participating children. However, we believe that such a presentation would turn out even more complex, and confusing for the reader, so we decided to present it as if it were a bone fide cohort study - with the added wrinkle that, due to very low number of participant numbers for most of the age groups, the effect is arguably more cosmetic than methodological.
Participants
The participants of this study were 23 Russian-Cypriot Greek simultaneous bilinguals, 11 girls and 12 boys. Their age ranged from 3 to 11 years, though participants numbers for most of the age groups were very small indeed: 3;1 (N=1), 4;8 (N=2), 5;0–5;6 (N=5), 6;0–6;11 (N=9), 7;11 (N=2), 9;5 (N=2), 10;11 (N=1), and 11;4 (N=1). At the time of testing, they attended kindergarten, pre-primary, and primary school classes. All participating children came from mixed-marriage families, with a Greek Cypriot father and a Russian mother, in a middle-class setting. They were randomly recruited in urban and rural areas of the Larnaca and Nicosia districts of Cyprus.
Materials and procedure
The Multilingual Assessment Instrument for Narratives (MAIN), a narrative tool developed within COST Action IS0804, was used in order to elicit stories from bilingual children (Gagarina et al., 2012; 2015). The MAIN consists of four comparable six-picture stories. Two of the picture sequences were used for the telling condition and another two for retelling. Each sequence consists of six coloured pictures without text.
For the telling mode, the Baby Goats and the Hungry Cat stories were chosen; children were asked to come up with and tell the experimenter a story based on the six pictures. For the retelling mode, the Baby Birds and the Naughty Dog stories were chosen; children were asked to first listen to the story told by the experimenter and then retell it. There was mutual sharing of the visual context and stimuli between child and examiner. During the testing both the child and the examiner could see the pictures.
Each child was tested individually in their home environment.
All participants were also tested on a battery of additional tests: the Diagnostic Verbal IQ Test (Stavrakaki & Tsimpli, 2000), adapted to CG from the Standard Modern Greek original (Theodorou, 2013), the Russian Proficiency Test for Multilingual Children (Gagarina, Klassert, & Topaj, 2010), and several tasks assessing executive functions (digit span test, word span test, fluency test, Raven’s Coloured Progressive Matrices). A parental questionnaire focusing on participants’ socio-economic and family language background was also used (Gagarina et al., 2010).
Data analysis
All data were recorded, transcribed, and analysed in terms of story structure, structural complexity per episode, and internal state terms. The analysis of story structure includes setting, mental state as initiating event, goal, attempt, outcome, and mental state as reaction (3 episodes in total). Structural complexity per episode (episode completeness) focused on whether children used Goal-Attempt- Outcome (GAO) in every episode. Internal state terms denote the types of mental state terms used by bilingual children in their narrative production.
Since Premack & Woodruff (1978), mental or internal state terms (ISTs) have been argued to relate to theory of mind and cognitive abilities. There are different types of ISTs, which can be classified, for example, into six categories (for further discussion, see Klop, 2011; Gagarina et al., 2012; 2015):
perceptual verbs (such as see, hear, feel, smell), physiological adjectives (thirsty, hungry, tired, sore), predicates expressing consciousness (alive, awake, asleep), emotional adjectives (e.g., sad, happy, angry, worried, disappointed), mental predicates (e.g., want, think, know, forget, decide, believe, wonder, have/make a plan), and verbs of saying or ‘linguistic verbs’ (e.g., say, call, shout, warn, ask).
128
Results
Overall, it was found that with regard to narrative abilities (macro-structure: story structure, structural complexity, and ISTs), the bilingual children performed similarly across their two languages (slightly better for CG). Their performance was also higher on the retelling than the telling condition. With respect to story structure for the four stories, within and cross-language comparison showed that the bilingual Russian-CG children performed better in retelling than in telling. For the telling mode, there was no crucial difference between the Russian and CG productions of the two stories, Baby Goat and Hungry Cat. For retelling of the Baby Bird story, the bilingual children had a slightly better production in CG than in Russian, but the opposite held for the Naughty Dog story. Table 1 presents the overall scores (and percentages) for the quantitative measurement of story structure per story for the combined group of 23 participants (with the proviso that it includes productions from individuals aged 3 to 11 lumped together); the MAIN manual provides a maximum score of 17 for each story.
Table 1. Story structure (telling vs. retelling)
Story structure CG Russian
Baby Goat (telling) 141 (36.1%) 143 (36.6%) Hungry Cat (telling) 144 (36.8%) 145 (37.1%) Baby Bird (retelling) 188 (48.1%) 169 (43.2%) Naughty Dog (retelling) 180 (46.0%) 186 (47.6%)
Table 2. Story complexity (telling vs. retelling) Story/Mode
Struct.
Complex.
(TOTAL) AO/AA GA/GO GAO
CG Baby Birds story structure 121 32 32 57
Russian Baby Birds story structure 99 37 38 24
CG Naughty Dog story structure 102 36 24 42
Russian Naughty Dog story structure 79 24 28 27
RETELLING TOTAL 401 129 122 150
CG Baby Goat story structure 76 49 18 9
Russian Baby Goat story structure 75 45 15 15
CG Hungry Cat story structure 62 18 20 24
Russian Hungry Cat story structure 61 13 18 30
TELLING TOTAL 274 125 71 78
CG TOTAL 361 135 94 132
Russian TOTAL 314 119 99 96
Proceedings ISMBS 2015
129
Table 3. Story structure, structural complexity, and internal state terms (means)
VARIABLES TELLING RETELLING
Language Age N Story
structure
Structural complexity
Internal state terms
Story structure
Structural complexity
Internal state terms Mean
CG 80
23 6.2 3 3.8 8 4.8 5.5
SD 23.6 2.4 1.5 2.2 2.6 1.5 2.9
Mean
Russian 80
23 6.2 2.9 4.1 7.7 3.8 5.9
SD 23.6 2.6 1.4 1.9 2.6 1.5 2.1
Table 4. Internal state terms (telling vs. retelling) INTERNAL STATE TERMS (TELLING/PRODUCTION) Language Perceptual
state terms
Physiological state terms
Consciousness terms
Emotion terms
Mental verbs
Linguistic verbs
Total Cypriot Greek 68 0 0 29 61 15
Mean 2.95 0 0 1.26 2.65 0.65
SD 2.28 0 0 1.83 2.08 1.43
Total Russian 84 1 0 27 62 12
Mean 3.65 0.04 0 1.17 2.69 0.52
SD 2.65 0.20 0 1.33 1.86 0.89
INTERNAL STATE TERMS (RETELLING/PRODUCTION) Language Perceptual
state terms
Physiological state terms
Consciousness terms
Emotion terms
Mental verbs
Linguistic verbs
Total Cypriot Greek 79 4 12 67 58 33
Mean 3.43 0.17 0.52 2.91 2.52 1.43
SD 2.29 0.38 2.5 2.72 1.92 1.30
Total Russian 89 9 0 55 73 46
Mean 3.86 0.39 0 2.39 3.17 2
SD 2.83 0.72 0 2.03 2.55 1.70
The analysis of structural complexity, qualitative organisation of episode structure, and macro- proposition that compose the plot (GAO) showed that the bilingual Russian-CG children had more structural complexity in retelling than in telling (total structural complexity, GA/GO, GAO); they also performed better in CG than in Russian. See Table 2 and Table 3 for the results, respectively.
The reader will recall the small participant numbers across too many age groups. Consequently, the results depicted in Tables 2 and 3 are obviously not very helpful, only indicative: They correspond to the mean performance of all children combined, from the single 3-year-old to the single 11-year-old.
The above-mentioned different single-case studies approach might work better, thus presenting each child’s scores individually. However, current space restrictions do not allow such a detailed analysis, which is why we restrict ourselves to reporting the data as if collected from a comparable cohort.
The analysis of the data further showed that the bilingual children used more ISTs in retelling than in telling. Specifically, they used more perceptual state terms, emotion terms, and mental verbs rather than psychological, consciousness terms, and linguistic verbs, as shown in table 4. According to a paired samples t-test, there is a statistically significant difference between ISTs in CG telling mode
130
and CG retelling move (t(21)=4.577; p=.000) as well as between ISTs in Russian telling and Russian retelling modes (t(21)=4.902; p=.000).
These findings are somewhat more meaningful than the above. For starters, regardless of age (from 3 to 11), certain ISTs were never used, others very rarely, and yet others more frequently. The table also highlights similarities and potential differences between the two languages. In order to explore this further, however, an individual approach would have to be taken, which cannot be done here.
Please see Appendix 1 for the bilingual children’s individual productions on story structure, Appendix 2 on structural complexity per episode, and Appendix 3 on total number of ISTs in tokens.
As expected, the bilingual children’s narrative abilities in CG improve with age, although the numbers of participants in each age group are too low to generalise this (beyond 5- and 6-year-olds, perhaps);
there is no such clear picture for Russian. Table 5 presents the raw scores for each age group and as such is more informative than Table 1 above. Note that all children were tested only on production and not comprehension. As said above, the maximum score for story structure is 17, while for internal state terms, the total number of IST tokens is counted. In terms of structural complexity per episode, within the Goal-Attempt-Outcome structured episodes, the scores represent how often a participant produces partial event sequences (AO, AA), incomplete episodes (GA/GO), and the targeted fully complete episodes (GAO); consequently, children get 1 point for each AO/AA, 2 points for each GA/GO and 3 points for each GAO (Gagarina et al., 2015).
Table 5. Story structure, structural complexity, and internal state terms (Age: telling vs. retelling)
VARIABLES TELLING RETELLING
Age Language Age
months N Story structure
Structural complexity
Internal state terms
Story structure
Structural complexity
Internal state terms 3-year-olds CG
47 1
3.5 2 3.5 4 3.5 0.5
Russian 3 2 2 4.5 3 5
4-year-olds CG
56 2 6 2.5 4.2 7.5 5.5 6
Russian 6.5 2.7 3 7.2 3.2 6.5
5-year-olds CG
63.8 5
3.8 1.7 2.5 6.3 4.5 3.5
Russian 2.8 1.4 3.3 5 3.1 4.5
6-year-olds CG
75.4 9 6.7 3.2 3 8.2 4.6 5.3
Russian 7.8 3.3 4.2 8.4 4 5.7
7-year-olds CG
95 2 8 4.2 6 10.2 5.7 7
Russian 6.7 2.7 6 8.25 4 6.7
9-year-olds CG
113 2 8.5 4 7.7 10.2 5 11
Russian 8 5.5 5.7 10.5 4.5 8
10-year-olds CG
131 1 8 5.5 6 11 7 8
Russian 8.5 5 6.5 12 7 9
11-year-olds CG
136 1 6.5 2 3 7 4 4
Russian 5 1.5 2 8 3.5 5.5
Since this table breaks the participants down into age groups, there is some comparability for each row. We thus yield a first indication of what age-related differences in performance could look like.
Due to the low number of participants, it does not make sense, though, to dwell on this further; more data from more participants are needed for each age group, except perhaps the 5- and 6-year-olds:
Proceedings ISMBS 2015
131
There is a noticeable level of improvement from age 5 (N=5) to age 6 (N=9) for each level of macro- structural analysis as well as for retelling over telling, and for both languages.
Next, we compare our data from bilingual Russian-CG children with available data on monolingual Russian- and monolingual/bilectal CG-speaking children (Gagarina et al., 2012), both with typical language development (TLD) and with specific language impairment (SLI). Looking at the narrative productions in both modes (telling and retelling), monolinguals outperform their bilingual peers mainly in story structure, as shown in Table 6.
Table 6. Bilingual TLD vs. monolingual TLD and SLI (telling vs. retelling)
TELLING/PRODUCTION RETELLING/PRODUCTION
Language Age N Story
structure
Structural complexity
Internal state terms
Story structure
Structural complexity
Internal state terms Monolingual children with TLD (Gagarina et al. 2012: 96)
Cypriot Greek 79.8 6 5.0 1.5 5.5 5.2 0.3 6.1
Russian 68 15 7.3 N/A 1.3 14.8 N/A 2.7
Greek 73.0 5 9.8 1.8 5.9 11.5 1.4 6.6
Bilingual children Cypriot Greek
63.8 5 3.8 1.7 2.5 6.3 4.5 3.5
Russian 2.8 1.4 3.3 5 3.1 4.5
Cypriot Greek 75.4 9 6.7 3.2 3 8.2 4.6 5.3
Russian 7.8 3.3 4.2 8.4 4 5.7
Cypriot Greek 95 2 8 4.2 6 10.2 5.7 7
Russian 6.7 2.7 6 8.25 4 6.7
Cypriot Greek 113 2 8.5 4 7.7 10.2 5 11
Russian 8 5.5 5.7 10.5 4.5 8
Monolingual children with SLI (Gagarina et al., 2012: 96)
Russian 68 9 6.7 N/A 1.9 6.7 N/A 2.1
Greek 100.6 18 3.9 0.4 2.8 5.8 1.0 5.5
Note that we only use a subset of the bilingual participants (total N=18), somewhat matching the children’s age from the studies compared to.
The comparison of our data from bilingual Russian-CG children with available data on monolingual Russian children (Gagarina et al., 2012), both with TLD and with SLI, showed that monolingual Russian children with TLD outperform their bilingual peers on story structure (telling and retelling), while the bilingual Russian-CG children scored higher on internal state terms (telling and retelling).
Bilingual Russian-CG children were closer to monolingual Russian children with SLI in terms of story structure, but they were better on internal state terms. This is shown in Table 7. Again, we only employed a subset (total N=14) to match those children we have data for.